app.py

import os
import torch
import torch.nn as nn
import numpy as np
import clip
import gradio as gr
from PIL import Image
from huggingface_hub import hf_hub_download

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
print(f"[info] device: {DEVICE}")

print("[info] Loading CLIP ViT-L/14 ...")
clip_model, preprocess = clip.load("ViT-L/14", device=DEVICE)
clip_model.eval()

print("[info] Downloading aesthetic-classifier checkpoint ...")
ckpt_path = hf_hub_download(
    repo_id="purplesmartai/aesthetic-classifier",
    filename="v2.ckpt",
)
checkpoint_data = torch.load(ckpt_path, map_location=DEVICE)
state_dict = checkpoint_data["state_dict"]
state_dict = {k.replace("model.", ""): v for k, v in state_dict.items()}

aesthetic_model = nn.Sequential(
    nn.Linear(768, 1024),
    nn.ReLU(),
    nn.Dropout(0.5),
    nn.Linear(1024, 512),
    nn.ReLU(),
    nn.Dropout(0.3),
    nn.Linear(512, 1),
).to(DEVICE)
aesthetic_model.load_state_dict(state_dict)
aesthetic_model.eval()
print("[info] Model ready.")


@torch.no_grad()
def get_score(image: Image.Image) -> float:
    t = preprocess(image.convert("RGB")).unsqueeze(0).to(DEVICE)
    feat = clip_model.encode_image(t).cpu().numpy().astype("float32")
    norm = np.linalg.norm(feat, axis=1, keepdims=True)
    feat = feat / np.where(norm == 0, 1, norm)
    return aesthetic_model(torch.tensor(feat, device=DEVICE)).item()


def raw_to_pony(raw: float) -> int:
    return int(max(0.0, min(0.99, raw)) * 10)


COLOURS = [
    "#c0392b", "#e74c3c", "#e67e22", "#f39c12", "#d4ac0d",
    "#27ae60", "#1e8449", "#148f77", "#0e6655", "#0a4f42",
]


def build_html(raw: float) -> str:
    pony   = raw_to_pony(raw)
    colour = COLOURS[pony]

    # Two rows of 5 so the grid never overflows
    rows = []
    for row_start in (0, 5):
        cells = ""
        for i in range(row_start, row_start + 5):
            active = i == pony
            bg     = COLOURS[i] if active else "rgba(255,255,255,0.07)"
            border = f"2px solid {COLOURS[i]}" if active else "2px solid rgba(255,255,255,0.12)"
            weight = "700" if active else "400"
            scale  = "scale(1.08)" if active else "scale(1)"
            opac   = "1" if active else "0.5"
            cells += (
                f'<div style="background:{bg};border:{border};border-radius:8px;'
                f'padding:9px 4px;text-align:center;font-size:.78rem;font-weight:{weight};'
                f'color:#fff;transform:{scale};opacity:{opac};transition:all .2s;'
                f'user-select:none;white-space:nowrap;">'
                f"score_{i}</div>"
            )
        rows.append(
            f'<div style="display:grid;grid-template-columns:repeat(5,1fr);gap:5px;margin-bottom:5px;">'
            f"{cells}</div>"
        )

    bar_w = min(max(raw, 0.0), 1.0) * 100
    return f"""
<div style="font-family:'Inter',sans-serif;padding:8px 0;">

  <div style="text-align:center;margin-bottom:18px;">
    <div style="display:inline-block;background:{colour};color:#fff;border-radius:12px;
        padding:12px 32px;font-size:1.9rem;font-weight:800;letter-spacing:.04em;
        box-shadow:0 4px 20px {colour}66;">score_{pony}</div>
    <div style="color:#aaa;font-size:.82rem;margin-top:7px;">
      raw: <code style="color:#ddd">{raw:.4f}</code>
    </div>
  </div>

  {"".join(rows)}

  <div style="background:rgba(255,255,255,.1);border-radius:6px;height:7px;overflow:hidden;margin-top:8px;">
    <div style="width:{bar_w:.1f}%;height:100%;
        background:linear-gradient(90deg,#c0392b,#f39c12,#27ae60);
        border-radius:6px;"></div>
  </div>
  <div style="display:flex;justify-content:space-between;font-size:.7rem;color:#666;margin-top:4px;">
    <span>score_0</span><span>score_9</span>
  </div>

</div>"""


def classify(image):
    if image is None:
        return "<p style='color:#888;text-align:center;padding:40px 0'>Upload an image to score it.</p>"
    return build_html(get_score(image))


with gr.Blocks(
    title="Aesthetic Classifier — PurpleSmartAI",
    theme=gr.themes.Soft(primary_hue="purple"),
    css=".gradio-container{max-width:860px!important;margin:auto}"
        " #title{text-align:center} #sub{text-align:center;color:#888;font-size:.9rem;margin-bottom:1.4rem}",
) as demo:
    gr.Markdown("# 🎨 Aesthetic Classifier", elem_id="title")
    gr.Markdown(
        "CLIP ViT-L/14 regression model by **PurpleSmartAI** for Pony V7 captioning. "
        "Outputs a **score_0…score_9** tag used directly in training captions.",
        elem_id="sub",
    )
    with gr.Row():
        with gr.Column(scale=1):
            img_input = gr.Image(type="pil", label="Input Image", height=340)
            run_btn   = gr.Button("✨ Score image", variant="primary", size="lg")
        with gr.Column(scale=1):
            out_html = gr.HTML(
                value="<p style='color:#888;text-align:center;padding:40px 0'>"
                      "Upload an image to see its score.</p>",
            )
    gr.Markdown(
        "---\n**Model:** [`purplesmartai/aesthetic-classifier`]"
        "(https://huggingface.co/purplesmartai/aesthetic-classifier)"
        " · **Backbone:** OpenAI CLIP ViT-L/14"
    )
    run_btn.click(fn=classify, inputs=img_input, outputs=out_html)
    img_input.change(fn=classify, inputs=img_input, outputs=out_html)

if __name__ == "__main__":
    demo.launch()